rollup-boost and op-rbuilder have grown beyond just Flashbots — they’re becoming critical pieces of infrastructure for multiple other teams and ecosystems. To support that, we’re committing to making their development fully open and transparent. As a first step, we’re starting to share recordings from our internal demo days! Over time, we’d love to open up these calls to the public and bring in more community contributors, and eventually do the entirety of ideation and development in public channels. But alas, we must move one step at a time to not overwhelm ourselves.
A summary of the demos for this inaugural week:
rollup-boost: the start of our new docs page as well as various tools for testing locally and benchmarking.
op-rbuilder: flashblocks demo on local environment.
Builder Playground v2: The first demo, a local end-to-end testing environment for OP Stack block builders that extends our original L1 testing environment.
contender: Demo of a new univ3 scenario, more complex spamming strategies, as well as a glimpse at an expansion into benchmarking Engine API calls.
chain-monitor: devops tool for monitoring block builder and batcher health.
general: demo of using LLMs to keep up with OSS Layer 2 developments.
In this demo I show how a Rollup-Boost-enabled block builder can serve 250ms Flashblocks. Using a local OP Node emulator, we show how Rollup-Boost queries the external builder for blocks and receives the “preconfirmations.”
Contender is a high-performance Ethereum network spammer and testing tool designed for benchmarking and stress-testing Ethereum clients and networks. In this demo I showcase running a new spam scenario which under the hood spams multiple UniswapV3 pools, and then use contender to generate a performance report that visualizes storage slot access as a heatmap showing hotspots of activity, gas usage spikes, and transaction inclusion times.
The first demo of Builder Playground v2, a local end-to-end testing environment for OP Stack block builders that extends our original L1 testing environment.
Builder Playground now supports two recipes - L1 and OP Stack. The tool generates useful artifacts including configuration files, logs, and graph visualizations showing service connections, making it much more efficient for quick feedback loops during development compared to previous deployment methods.
I demonstrate running a basic Contender scenario to put load on the Unichain Experimental Testnet using the Uniswap v3 scenario at 20TPS for 8 seconds.
After running the transactions, we generate a performance report that provides metrics about the chain’s response to the load, allowing us to assess how well the experimental testnet handled the transaction volume.
I demonstrate op-rbuilder with devnet parameters in two modes: default mode cycling through empty payloads, and with Contender flag enabled to maximize block utilization.
With Contender active, the system spams the chain with as many transactions as possible without exceeding the gas limit.
This demo show how we use mdBook to manage Rollup-Boost documentation in our repository. I demonstrate adding an architecture overview by incorporating the system workflow from the README into the docs, which updates in the browser and properly renders diagrams from the markdown content. Evenutally we will use this configuration for our official Rollup Boost docs to minimize overhead in maintenance.
In this demo I’ll show some recent updates to Builder Playground v2 by deploying an OP Stack and connecting an external op-rbuilder and dumping TCP flows to manually inspect traffic.
The main focus of the demo is a new inspector tool I’ve developed that parses TCP flow data and presents it as readable, structured log outputs instead of raw data dumps. We’re evaluating whether to integrate similar functionality directly into the reth client.
This demo is showcasing chain-monitor, a tool we built to determine if blocks on OP Stack chains are built by our builder rather than the fallback builder. The system works by identifying a “magical transaction” our builder includes in every block, allowing us to track metrics like landed versus missed blocks.
Chain-monitor also monitors crucial wallet balances (batcher, proposer, builder) and measures transaction inclusion times, providing dashboards and alerts to help operators maintain optimal chain performance.
There’s too much information and too many Telegram chats for OSS development. In this demo I demonstrating two tools I’ve built to help keep up: a GitHub repo scraper and a Telegram group chat scraper. These Python-based tools generate static HTML weekly reports with metrics on PRs, issues, response times, and contributor activity, along with AI-generated summaries of key discussions and decisions.
The goal is to eventually integrate these into GitHub Actions for automated updates, providing a low-bandwidth way to track developments across multiple repositories and communication channels.
This demo shows a sneak peak of how we’ve extended Contender to interact directly with the Engine API, enabling faster benchmarking than typical devnets. By running Reth with a devchain and without its own consensus layer, I can use Contender’s new JWT and FCU arguments to directly access the Engine API. This allows us to send transactions to the mempool and then immediately instruct the node to build blocks, either in one-second intervals (TPS mode) or batch-by-block (TPB mode), potentially offering significant speedups for CI testing.
